Imagine you are a PM in Amazon. How would you reduce the rate of returns?

Assume current rate of returing orders is 12% and you have to bring it down to 8%.  Based on your analysis you found the most common reason for returning is ” Product doesn’t match description”.Assume total different kinds of products listed = 10,000.

How will you proceed from here ?

  Amazon
  Stripe
Add Your Answer
Answers (5)

C: Clarify the situation

We have 10,000 items that have a total return rate of 12%. First thing I would do is graph the distribution of return rate per type

There are two ways this can come out

1. The distribution is even among all 10,000 items

2. The distribution is biased to a specific type of items.

 

 

Regarding 1. This would mean that the categorization is too granular. I would further cluster the 10,000 items to bigger clusters until we see a clear #2 pattern.

 

I: Identify the customers

I would take it General, and Market Prioritized

Generally, what kinds of customers return the most?

1. High return for some reason or another customer?

2. Try on stuff to see fit customer: Customers who purchase different sizes of the same item and return the wrong sizes?

3. Use once and return customer: These customers are the ones that return after using the item once.

 

I would look at the distribution of these customers across the general 10,000 types, and across the largest cluster identified in the step above. Compare them. If they are starkly different. The two approaches that we would want to look at is

 

1. Do we want to target the segment that has the highest return rate?

2. Do you want to target the type of customer that has the highest return rate?

Imagine you are a PM in Amazon. How would you reduce the rate of returns?

I will apply estimation, product strategy, and metrics to solve such questions.

 

Let’s consider daily shipping = 500 products of 10000 product listings

Current returing rate = 12% = 500*.12= 60 (Actuals)

Expected returning rate= 8%= 500*.08=40 (Targeted)

Assumptions: let’s assume we will achive this target in a quater.

Interviewer: okay!!

 

I’ll rule out all other assumptions for the return and solely assume 60 returns are happening due to 1 single reason which is “description mismatch”.

 

Define the problem statement :

How might we reduce the report of the products from 12% to 8% that is caused by product description mismatch reason?

Let’s analyze the top pain points of the customers that happen during their shopping experience.

The typical ( easiest form) customer journey looks like this-

login-> Discovery-> checkout->payment-> confirmation-> Delivery

The main pain point occurs during the product discovery stage when a customer might make a wrong decision and add the product to the cart.

Potential pain points during the discovery phase :

 

1)Misleading size,

2) Misleading colour

3) Misleading design

4) Misleading reviews/rating

5) Inaccurate images

6) Wrong/misleading descriptions

6) Unable to identify sellers’ service & past performance.

 

Note: I will prioritize all the above pain points to fix to achieve the target.

Potential solutions:

let’s categorize the solution in two ways-

1) Operational enhancement -> for Sellers’ persona

2) Product experience -> Buyers’ persona

 

Operational enhancement :

1) Ensures robust seller verification and integrity check based on their history. [Mid effort, high business impact]

2) Enable authentic seller rating/ranking schemes based on historical service [ Mid effort, mid-business impact]

3) Endorsing sellers based on their selling history and customer reviews. [ easy effort, mid-business impact]

Experience improvement :

4) Highlight the label of authenticity for product sellers so that end-users understand the seller’s services before making a decision.

The attribute of authenticity can be-

Prime status, qualified seller, success service rating, and highly rated seller’s products are showcased as trending. [ easy to deliver, high business impact]

5) Display 360 experience, display accurate size, colour, clear description, and how the product looks when used. [ mid effort to deliver, mid business impact]

6)No out-of-stock/ sold-out products display [easy to deliver, high business impact]

 

Metrics:

In this case, my target is a little aggressive to bring down the return rate from 12% to 8% in a quarter. Hence, I will track the measures week, monthly and quarterly basis so that we can track any deviation sooner than later.

 The week on Week tracking: lead metrics/supporting metrics

1-x% reduction in the return of items

2- X point improvement in User feedback ( csat ) on categories-

Quality, delivery & packaging

 

Month on month: Tracking metrics

% no of sellers acquired an improved rating

% no of sellers qualified for prime

% no of existing/new sellers qualified as top partners

 

Quater on Quater: Lag metric

1- % target achieve to reduce the return from 12% to 8% in a quarter.

This metric will show that part of the target achieved either 100% or maybe less or more.

 

Some strategic aspects:

Run an A/B test for an experience enhancement to verify if the solutions proposed are working or not.

The first step here will to segment return orders  by product category and supplier. If its product category then is this return expected? For example for clothing return rates could be high as people might try and may have size issue or just don’t like it on trying as much.  If this is not expected, then we should analyze user review data  (run user studies) to understand the root cause.

1) Is it due to misunderstanding from users? Mismatch in expectations.

2) Is there a decription understanding issue like language, gaps

3)Is it supplier’s fault i.e they did it on purpose to boost their sales.

As one of the solution,we can introduce a checkbox feedback from user on whether product recived matches with description? If no, then for that supplier we can add negative score to penalize them in the ranking stack. We should make the number of votes on checkbox collected as feedback visibile to users to help them make informed decision.

If there are repeated complains of user on supplier then amazon should add the supplier in blocklist. If it’s language and understanding issue then we can have suppliers fix the description.

A more advanced solution would be to train  model to scan review data and provide score on how closely does the product matches the description by scanning through description keywords in user’s feedback or positve feedbcak from users.

I would imagine that the goal of the problem is to see how you think through a problem to identify a root cause, and then implement solutions considering trade-offs.

Understand the problem first to get to the root cause:

Are these products being returned in a particular category i.e apparel versus watches? This might give some clues about how the product descriptions or SKUs are being listed on the website today.

Are these FBA or are these shipped directly by merchants to customers? How much can Amazon control in this process? Is there an error in how the products are being stocked in the warehouse or is there a problem with how the seller has separated SKUs?

Is this problem pervasive across sellers in a single category or multiple sellers in multiple categories?

Are these products being shipped from a new fulfillment center ? Was there a turnover in staff recently in these fulfillment centers?

Is this a new product category for Amazon? Are the product descriptions accurate?

Solutions

1. If the problem is specific to a category i..e tee shirts, see whether the product detail pages for these items identify the right item to ship based on Color (blue, black, white) , Size (S,M,L) and any other factors that can multiply the inventory.

2. If the problem is specific to a seller, investigate Do items have duplicate bar codes even though they are completely different? check if the seller is new or requires better quality control.

3. If the problem is with a specific fulfillment center and not a specific type of product or seller, then identify where in the FC the problem occurs.

4. Is the seller shipping on their own? Do they need to be shut down for breaching customer trust?

This problem solving interview question calls for an analysis of various numbers.

Total Number of Products: 10,000 (Given)

Total Number of orders shipped: 1,000 (Assumption)

Total Number of orders returned: 120 (12% of 1,000)

Probable Reasons for Return:

  • Broken Item -20(16.6% of Total returned)
  • Late Delivery -20(16.6% of Total returned)
  • Mistakenly Ordered -20(16.6% of Total returned)
  • Description Mismatch -60(50% of Total returned)
    • Seller
      • Intentional
        • Knowningly uploads wrong Information
        • Knowningly uploads less Information
        • Knowningly uploads confusing Information
      • Unintentional
        • Unknowningly uploads wrong Information
        • Unknowningly uploads less Information
        • Unknowningly uploads confusing Information
    • Buyer
      • Less Information
      • Doesn’t read Information
      • Misunderstand Information

Information Based upon which Buyer makes buying decision:

  • Pictures
    • Color
    • Shape
    • Size
  • Description
    • Material
    • Color
    • Size
    • Price
  • Reviews

Our Objective is to reduce returns by 4%(12%-8%) which leads to only 80 returns from earlier value of 120. Since “Wrong Description” accounts for 50% of returns. We need to reduce “Wrong Description” returns from earlier value of 60 to 20 i.e. 66% percent of reduction.

 

Problem Statement: Buyers return items on Amazon because buyer doesn’t make well informed decision at the time of purchase. Information is either lacking, confusing, or wrong for buyer.

 

Solutions: As part of solution, we need to make sure that buyer has all the information required to make right decision and make sure buyer is utilizing that information and making right decisions.

 

  • Less Information: What more information does a buyer need to make a right buying decision.
    • Picture: 3D Imaging.
    • Videos: 3D Video with sound description.
    • Description: Material,Color,specifications etc tagging and comparison with 3D images. Standard scale to compare height, weight, width etc.
    • Written+ Audio+Video Reviews: Amazon points for product reviews and feedback. New review system for customers which can have audio,video, and text functionality. Existing customer can make 30 sec-1 min videos or audios.
    • Chat: Prospective customers can chat with existing buyers. Algorithm will match customers so as to maintain high customer experience for both prospective and existing customer.
  • Doesn’t read Information: How can we make sure buyer reads all the information about product.
    • Mutilingual: NLP will transform reviews, description in language of choice for buyer.
    • Picture: People tend to learn more through pictures so increasing #pictures may help.
    • Description: Change description to tags instead of written text.
    • Reviews: Sort reviews with negative and positive comments.
    • Scroll Notification: Record scrolling sessions to remind buyer to spend more time on reading description.
  • Misunderstand Information: Make sure buyer understands correctly about product.
    • Picture: Scaled images to highlight specifications of product.
    • Description: 3D imaging can help reduce errors around specifications like height, weight etc.
    • Reviews: Feature through which reviews can alter descriptions. A well written review can act as the secondary product description.

 

Criteria 3D Imaging Audio/Video Reviews Chat Multilingual Scroll Notification
Value to Customer H H H M L
Value to Platform H H M M L
Resource Investment M H M M L
Time Investment H M M M L

 

I would prioritize following three features over others.

  • 3D Imaging
  • Audio/Video Reviews
  • Chat

 

Metrics

  • Business Metrics
    • % Returns
    • Sales Conversion
  • Product Metrics
    • Feature Utilization (#times feature used)
    • Feature Penetration (#times feature used per product)
    • Feature Spread (#products on which feature used)
    • Return-Feature Usage Ratio
  • Ancillary Metrics
    • AOV
    • Retention Rate
    • Repeat Purchase Probability